So, what is computational social science then? I believe that no definitive answer exists yet and may not ever emerge. Rather, different communities have their own perspective on this topic. None of them is wrong. Instead, they highlight different pieces of the puzzle.
However, what has remained hidden from all perspectives above is the role of social science theory in the research. By this I do not mean if the research is deductive (using existing theories to study a phenomenon) or inductive (using a phenomenon to propose novel theoretical forms). Nor do I seek to question what `theory' is and characteristics of good theory in terms of falsification or conceptual clarity. Within this book, social science theory refers to the collective knowledge developed and conceptualised in social scientists over decades. My question on the role of social science theory in computational social science is if computational social science should utilise and, in the long run, contribute to social science theory.
As a social scientist, I believe that the answer is yes. In the beginning of this chapter, I introduced discussion supporting this approach (Wallach, 2018; Grimmer, 2015). However, the environment is changing. As Grimmer (2015) suggests, nowadays everyone can be a social scientist if they want to. Some even argue that this may lead to a crisis in academic social science fields (Savage and Burrows, 2009). However, the situation may not be that bad. On a more positive tone, Grimmer (2015) argues that social scientists have skills that other people in the domain do not have, such as causal analysis. Wallach (2018) highlights that social scientists seek to explain and not only predict. According to her, this creates different demands towards computational tasks. Beyond these aspects, I argue that there is a lot more that social scientists can bring to computational social sciences.
Already the traditional education has developed valuable skills to help in computational social science work. For example, skills to reflect social processes that produce data analysed are critical for research. Big data are never context free (Boyd and Crawford, 2012). Early on, different forms of response bias are known by quantitatively oriented social scientists. Qualitative researchers similarly see the complex nature of social interaction and settings, which may help them to interpret and analyse these phenomena. There are many methods to address messiness of real data which are familiar to social scientists, but less known by computer scientists â like thinking about human-classified data and its accuracy (, ). Social scientists spend several years in their education to understand these aspects of research processes and how to design research and take challenges related to (social) data into account. (And they should be proud of that.)
Beyond these skills, social scientists are armed with years of social science theory. They have spent several years developing conceptual tools that help reflect social phenomena and direct investigations. Such tools and insights may help to pinpoint and reflect on core issues and engage in a deeper analysis of them. To illustrate the unique value of this work, consider supporting political engagement through technology. There is an extensive body of work in human-computer interaction that has sought to create new environments for voting and, thus, citizen participation (for review, see Nelimarkka, 2019). They have aimed to make voting `lightweight' (Taylor et al., 2012), `lower[ing] the barrier' to voting (Xu et al., 2017) and considering `cross-sectional citizens' who have limited abilities to participate (Fredericks et al., 2015). However, these works have not sought to deeply engage in political science, where it is well established that voting activity is dependent on several socio-economic factors (Nelimarkka, 2019). Thus, none of the papers makes an analysis on who uses the novel voting systems created. Here, I have chosen to illustrate my point through a human-computer interaction scholarship, as it claims to be truly collaborative, integrating holistic design, engineering and science, both social and natural (Shneiderman, 2016). (However, like always in science, these claims have been under critique and discussion has not yet finished (Marshall et al., 2017a,b; Nelimarkka, 2019).) A similar case could be done through many different cases, focusing on computer sciences and its subfield â or even focusing on boundaries within social sciences. Computational social science and other forms of interdisciplinary scholarly work are never easy as the work opens to various theoretical and methodological perspectives. In the best cases, these different perspectives can support the scholarship by pushing the boundaries of traditional disciplinary approaches.
Thus, I believe that social science contributions will become a critical component in computational social science. This said, we must acknowledge that a good amount of research addressing social phenomena can be published without anything resembling theory to social scientists. For example, the highly influential work on political polarisation in the United States blogsphere (Adamic and Glance, 2005) has only a handful of references to social sciences â and does not use the word polarisation anywhere in the article. Therefore, it does not deeply engage with existing social science knowledge about the topic but rather presents a phenomenon in a data-driven manner and with iconic visualisations. Such research is clearly increasing, as new research communities are engaging questions that have been traditionally in the domain of social sciences but come from various disciplinary backgrounds and traditions. The examples above demonstrated many such attempts: either scholars from a non-social science background applying and theorising about society or social scientists applying tools, approaches and ways of thinking originating from other disciplines.
However, unlike the sections above, I have not provided social theory-driven examples of computational social science. This is not because they do not exist. Rather, as social sciences take place in so many shapes and forms, it is easy to identify illustrative examples. Social science has many disciplines that use computational social science methods under a wide umbrella. For example, Evans and Aceves (2016) review extensively works in sociology that have applied computational text analysis, while Squazzoni et al. (2013) focused on simulation models. In the following chapters, example papers are discussed more extensively. We will return to the question of how to integrate computational components to social science research in Chapter 11. At that point, I hope you have gained additional tools and insights to have your own opinions and perspectives on this topic as well.